Alerts are sometimes not being fired when the logs meet the threshold criteria.

Aberta
#3,307 0 comentários 0 reações 0 responsáveis Ver no GitHub

Mantenedores costumam responder em até 1 dia

Ninguém assumiu esta issue ainda.

Avaliação

Dificuldade
4/5
Tempo estimado
3-5 dias
Facilidade para iniciantes
57/100
Tipo de issue
Bug
Clareza
Razoavelmente clara
Status de atividade
Ativa
Stack de tecnologia
clickhouse, typescript
Domínio
observability

Direção de pesquisa

Start by reading packages/api/src/tasks/checkAlerts/index.ts, especially getScheduledWindowStart, and compare the current Start offset behavior with the tooltip’s promise of delaying evaluation for ingestion lag. Confirm the intended window semantics and add or run tests for a 2-minute offset on a 5-minute interval; done means delayed evaluation checks the intended earlier window and alerts are not missed due to ingestion lag.

Escrita pelo modelo de indexação a partir do texto da issue.

Descrição

bug external Needs Feedback

What happened?

Hi, I'm using ClickStack and occasionally run into alerts that are not fired when they should be. This has only happened a few times over the past couple of months.

I think it may be caused by a log arriving after its window has been evaluated -- while investigating I found the Start offset setting under Advanced Settings on the alert, its tooltip says "Delays the start of each evaluation window by this many minutes. Useful when data is ingested with a lag," which sounds like exactly what I need.

However, from reading the source (getScheduledWindowStart in packages/api/src/tasks/checkAlerts/index.ts), I don't think it does this. With a 2 minute offset on a 5 minute interval, it shifts the window to e.g. 09:02:00 AM - 09:07:00 AM, rather than evaluating 09:00:00 AM - 09:05:00 AM two minutes later.

If my understanding is correct, I'm happy to put up a PR so the offset acts as a real evaluation delay for ingestion lag, as the tooltip describes, but wanted to check, can someone confirm the current behaviour isn't intended for a particular use case?

Steps to reproduce

  1. Create a saved query.
  2. Create an alert from that query which fires at >= 1 occurrences per 5-minute window.
  3. Sometimes the alert will not fire.

How are you running HyperDX?

ClickHouse Cloud.

Logs

The log's timestamp is 2026-10-06T23:09:46.990000000Z, which is only about 13 seconds before the window closed, again supporting the theory that it landed after the window was evaluated.

Image

However, the window is completely clear "OK" and no alert was ever sent.

Image

The timestamp on the log is 2026-10-06T23:09:46.990000000Z, so getting "close" to the window.

Linguagem predominante
TypeScript
Estrelas
10.1k
Forks
527
Merge médio
1d 20h
PRs com merge (30d)
186

Preparar o ambiente

Primeiros passos

  1. Leia a issue inteira e depois o guia de contribuição do projeto.
  2. Comente na issue dizendo que vai assumir — evita que duas pessoas façam o mesmo trabalho.
  3. Faça um fork do repositório e trabalhe em uma branch.
  4. Abra um pull request que referencie o número da issue.

Mais de ClickHouse/ClickStack

Todas as issues de ClickHouse/ClickStack

Issues semelhantes

Mais issues de TypeScript

Receba novas issues na sua caixa de entrada

Um resumo curto de issues do GitHub para quem está começando.